Relative Semantics Are More Real Than Absolute Semantics

Shared relative differences as foundational linguistic reference points

We walk outside into the summer air and feel the warm air blow onto your faces. All of us feel warmer. I say “What a hot night!” and everyone nods.

We do not all feel the same level of hotness—some of us would rather it was hotter and some would rather it was colder. But it is hotter than the air-conditioned building we left and it is one of the hottest nights in our little town.

Indeed, we have no really good way of talking about “how hot” one is feeling. We talk about these things in terms of function: I exclaim “It’s so hot you don’t need a jacket!”

Anne and Beth are wearing jackets.

Anne corrects me: “I still need a jacket, Crispy.”

“Well, of course you do, Anne. You’re always cold.” I reply.

I have factored-in Anne’s eternal coldness—the fact that it doesn’t change under these circumstances doesn’t really mean much to me. There are no jacketless nights for Anne in our town, so I don’t really think about her as playing the “Is it a jacketless night?” game.

Beth, on the other hand, just smiles. She doesn’t feel a need to correct me, because she knows how I would respond, and doesn’t feel like it would lead to a conversation of any importance.

Duke says, “Why do you always say it’s a jacketless night, when Anne’s always going to wear a jacket?” not because he doesn’t know why I said it, but because he’s indicating that I might as well play a game where everyone present is a participant. He’s indicating this without having to put me on the spot quite so hard, but enough that I feel a bit silly. This is “social pressure” and it’s a tool people use to show the kinds of games they want to play without insisting on them.

When I exclaim that you don’t need a jacket, I’m really making a point about how hot it is—one that degrees don’t really capture. This is why weather reports often have a “feels like” estimate, because the precise temperature is known to be a poor predictor of the actual experience of heat. Funnily, the “feels like” is also expressed in degrees. The idea, supposedly, is that this is being expressed in ideal conditions, e.g., without wind.

Even if the “feels like” estimate is accurate, people adjust differently to the same number. But our internal experience of the heat is mediated by how we react to it, which is why the first thing that jumped to my mind was the jacketlessness of the night. This observation was more than just personal—everyone felt the heat, because it was a relative change, but I wanted to express exactly what it meant to me and then see if I could socialize around the finer points of it, to see what kind of overlap with other sharing my experience of the world would lead to.

The resulting conversation ended-up being less about heat, and more about the meta-game of the claims I make and what social dynamics they create. I find this is often the case. That arguments about how to cut a cake, often have less to do with the cake, except as a lead-in. “I’ll cut the cake, but you can choose which piece.”

Relative semantics, like the increase in heat, are a vehicle for agreement in socialization. There are lots of relative differences we can agree on, where it’s difficult or impossible to characterize an absolute, e.g., how angry someone who left the party was.

Absolute semantics are only really possible once we reduce something down to observables. This is possible with lots of non-social stuff, because the context can be pinned down much easier. But even very simple social effects, like how spicy something is, are basically impossible to measure because the internal experience of it mediates the outward effect so much. We can measure someone’s biological reaction to capsaicin, but we can’t characterize what feelings they’re interpreting it as, we can’t use that to predict how spicy they like their food, and we can’t know how much they suffer spiciness out of machismo.

An absolute semantics of social effects is impossible with current tools, because most social effects are really about taking one path from a garden of forking paths, and what tipped things towards one option or another. Social games are reflexive, and what the participants think about each other matters a lot. For any fairly realistic social game, this information can’t be elicited. There is an absolute social calculus that could allow us to predict a person’s every actions, but short of simulation, our ability to elicit it will be very coarse for a very long time.

But we build up our models of other people by the information we get from relative semantics, and by using the “typical mind fallacy” judiciously. The typical mind fallacy is only a fallacy because a lot of suckers took it too far, and now we have to constantly argue about how far to take it instead of all the good stuff.

The good stuff is stuff like this: I can predict that any meme that’s used authentically when it’s getting popular will flip into being mostly sarcastic, and that the precise opposite will happen to initially sarcastic memes. It’s just a matter of pricing: when there are two natural interpretations and you dilute the value of one, the other is implicitly subsidized.

More on memetic subsidies soon.